Multiple classifiers systems with granular neural networks

Hybridization of neural networks and fuzzy sets has proved its efficiency in solving different pattern classification tasks, which led to the development of granular neural networks (GNNs). GNN works with the principles of granular computing and basically operates on granules of information. The present paper proposes an efficient multiple classifier system (MCS) framework with different guiding rules based GNNs. The performance of the proposed MCS is demonstrated and its superiority over individual GNNs is justified with remote sensing data for five land use/cover classes. Conventional back propagation algorithm is used to train the networks.